DE102009040145A1 - Method and device for stopping a manipulator - Google Patents

Abstract

The method involves monitoring an area (A0,A2) and braking a manipulator. The monitored area is variably specified during the operation on the basis of a braking distance (S-max) of the manipulator. The braking distance or a brake force distribution of the manipulator is preliminary simulated for different states. Independent claims are also included for the following: (1) a device for stopping a manipulator, particularly a robot; (2) a computer program that runs in a computer system; and (3) a computer program product with a program code.

Description

The present invention relates to a method and a device for stopping a manipulator, in particular a robot, as well as for determining a braking distance of the manipulator.

In particular in the context of safety-critical robot applications, it is known to specify protection areas, for example in a world coordinate system, into which the robot is not allowed to penetrate with links or reference points such as its TCP. If a zone violation is detected, the robot is safely shut down.

In this case, however, the robot can penetrate even further into the protected area due to its inertia and the dead times and inertias between range violation and response of the monitoring.

The DE 10 2008 013 431 A1 suggests to determine the maximum available delay time based on the distance of a detected person and the working range of a machine and the average walking speed of the person and to choose the braking course accordingly to shut down the machine from a hazard. The associated personal detection is very complex.

The object of the present invention is to improve the operation of manipulators.

This object is achieved by a method having the features of claim 1 or 2. Claim 9 provides a device, claim 10 or 11, a computer program or a computer program product, in particular a data carrier or a storage medium, for performing a method according to the invention under protection. The subclaims relate to advantageous developments.

According to a first aspect of the present invention, it is proposed to monitor an area and to brake the manipulator, preferably to shut it down safely when it is detected that it violates the monitored area.

The area can be specified for example in joint coordinates of the manipulator or world coordinates, for example a robot cell. For this purpose, it can be monitored, for example, on the basis of its joint coordinates, which may have been transformed into world coordinates, whether or not one or more members of the manipulator or manipulator-fixed reference points such as its TCP are within an inadmissible protection range or outside an admissible working range. Likewise, positions of such members or reference points can also be detected by a device, for example by means of electromagnetic radiation such as visible or UV light, radar, radio or the like, by one or more cameras detecting the pose of the manipulator or the distance of one or more tamper-resistant reference points is detected by ambient transmitters or receivers.

According to the first aspect of the present invention, the monitored area is now variably set during operation on the basis of a braking distance of the manipulator. In particular, it can be increased with longer braking distances and / or reduced in shorter braking distances. Preferably, the braking distance is determined during operation, for example cyclically, and the monitored area is updated accordingly.

The braking distance depends in particular on the kinetic energy of the manipulator to be dissipated, which in turn depends on its speeds, masses and moments of inertia, which in turn may vary depending on the moving load. In addition, the braking distance can also be extended or shortened by gravity, depending on whether it induces additional braking forces and moments or counteracts them in an accelerating manner. Furthermore, different braking force or torque curves due to tolerances, wear, heating and the like lead to different braking distances.

Accordingly, the braking distance after execution during operation on the basis of a model which describes the dynamics of the manipulator, in particular its weight and / or inertial forces, and / or Bremskraft- or -momentverläufe determined and determined accordingly the braking distance. If, for example, the moment of inertia J with respect to the vertical axis of a uniaxial robot with the joint angle q and its constant braking torque τ _{0 are} known, the braking distance s results on the basis of the dynamic model J × d ² q / dt ² = -τ (1) depending on its initial velocity ω by time integration s = (J × ω ² ) / (2 × τ ₀ ) (2).

For more complex manipulators such as six-axis industrial robots, the evaluation of an appropriate model for determining the braking distance s = f (q, dq / dt, L) (3) depending on the joint coordinates q, their first time derivative or joint speed dq / dt and load parameters L, which in addition to the load mass may include their moments of inertia and center of gravity, but hardly online during operation possible.

Therefore, according to a second aspect of the present invention, a subsequently explained method for determining a braking distance of a manipulator, in particular of a robot is proposed, which in particular for determining the braking distance based on a current state during operation, i. e. is suitable online. First and second aspects are preferably combined. However, as the above equation (2) shows, the first aspect can be utilized without the second aspect. Likewise, the second aspect can be used without realizing the first aspect, for example for path planning and optimization, in which a large number of simulation runs have to be carried out and braking distances have to be determined for different manipulator states.

According to the second aspect of the present invention, braking paths are simulated in advance, in particular numerically and / or empirically, for different states and / or braking force courses of the manipulator.

In this case, a state can be, for example, one or more positions and / or speeds, for example Cartesian positions or velocities of manipulator-fixed reference points in the world coordinate system, or joint coordinates, in particular joint angles or their time derivatives, load parameters such as in particular mass, moments of inertia and / or center of gravity of the load and the like. A numerical pre-simulation can be carried out in particular by the time integration of a dynamic model of the manipulator, as shown purely by way of example in Eq. (1) is an empirical through test runs and their evaluation.

A brake force curve can include, for example, the magnitude of a braking force as a function of the state, in particular the joint coordinates and / or their time derivatives, the time and / or other parameters such as static friction coefficients and sliding friction coefficients. In this case, anti-parallel braking force pairs, ie braking torques which act on an axis of rotation of the manipulator, are generally also referred to as braking forces for a more compact representation. A brake force curve can therefore, for example, in the form τ _i = μ ₀ × (dq _i / dt) / | dq _i / dt | + μ × (dq _i / dt) (4) be predetermined with the force acting on the axis i braking torque τ _i and the adhesion or sliding friction coefficient μ ₀ , μ <0. Instead of different braking force curves can be used for the variation of the states also the same brake force curve, for example, a minimum braking force curve for each axis, so that it is ensured that the actual braking force curve is higher, the actual braking distance is therefore always shorter than the simulated.

According to the invention, an upper limit for the braking distance is then estimated on the basis of these previously simulated braking distances. This takes place, in particular when combined with the first aspect, during the operation of the manipulator, preferably substantially in real time.

Due to the estimation of an upper barrier according to the invention, it is possible to dispense with the elaborate exact evaluation of the dynamic model or corresponding test runs, if the current braking distance has to be determined. For example, for a current state, one or more proximate simulated states may be used, and an upper bound for the current state may be estimated, for example interpolated or extrapolated, from the braking paths determined for them.

Already with reference to the highly simplified example with a vertical axis of rotation explained above, it can be seen that the braking distance includes, inter alia, inertia parameters of a load of the manipulator, in Eq. (1) depends on the moment of inertia J. These vary depending on the moving load, such as gripped payload, different tools, etc., sometimes strong.

Therefore, in a preferred embodiment of the present invention, the braking paths for different states of the manipulator for a reference load and / or Referenztraglastanordnung is simulated. Accordingly, a reference load can be predetermined, for example, by values for mass and moments of inertia, a reference load arrangement corresponding, for example, by the center of gravity of the load to the TCP or the axes of movement of the manipulator.

If an actual load in operation deviates too much from the reference load applied in pre-simulations, this distorts the estimated upper limit for the braking distance with the consequence that it becomes longer than necessary or too short. Therefore, in a preferred embodiment, a braking distance, which has been simulated in advance for a reference load and / or reference load arrangement, to an actual load and / or load assembly scaled to estimate the upper bound.

This can be done, for example, by simulating a scaling quantity on the one hand for the reference load and / or for the reference load arrangement, on the other for the actual load or load-bearing arrangement, and forming a scaling factor from the quotient of both scaling variables, with which the estimated braking distance is multiplied by way of example becomes.

Since in estimation based on a reference load and scaling of the braking distance with one or more scaling factors is not necessarily guaranteed that the braking distance at an actual load does not exceed the estimated upper bound, in a preferred development braking paths for different states of the manipulator additionally for simulates one or more of the reference load and / or Referenztraglastanordnung different payloads or load capacity arrangements in advance. If a longer braking distance is determined than that which results on the basis of the reference load and the scaling, the upper limit is increased accordingly. For this purpose, the braking distance, which was simulated for the reference load (arrangement) different load (arrangement), equally represented by the inverse of the or the scaling factors on the braking distance for the reference load (arrangement) and thus as a further simulation run with reference load ( arrangement).

For example, in a conventional six-axis articulated robot, a state vector of its joint angles and joint angular velocities already has 12 dimensions. Therefore, in a preferred embodiment of the present invention, a multi-dimensional state of the manipulator is mapped or transformed to a lower-dimensional, in particular scalar, size or overall speed. For this purpose, for example, a weighted sum of the amounts or powers of the joint speeds can be formed. Since the braking distance depends strongly on the kinetic energy of the manipulator and this results in articulated robots from the sum of the moments of inertia about the individual axes multiplied by half the squares of the corresponding joint angular velocities, in a preferred embodiment, the through the joint coordinates and their first time derivative given state is transformed to a scalar total velocity, which essentially corresponds to the kinetic energy in this state, wherein for the sake of simplicity, maximum values for load parameters such as moments of inertia and / or center of gravity distances can also be used.

In particular, when a multidimensional state of the manipulator is transformed to a scalar total velocity for estimation of the upper bound, the longest braking distance of all simulations for different states of the manipulator, which are transformed within a given discretization or tolerance to the same scalar total velocity Barrier can be selected for this scalar overall speed. The maximum braking distance can then be stored, for example, by a polygon over the scalar total speed.

In a preferred embodiment, the pre-simulations are performed type-specifically for various manipulators that differ in their configuration, such as axle arrangement, center distances, masses, moments of inertia, and other machine data. The determination of the braking distance for a specific manipulator then takes place on the basis of the brake paths simulated in advance for this type of manipulator.

Further advantages and features emerge from the subclaims and the exemplary embodiments. This shows, partially schematized:

1 . 2 : Protection areas of a robot, which are given according to an embodiment of the present invention; and

3 : the sequence of a method according to the invention for specifying the protection zones in 1 . 2 ,

For a better overview, the present invention is explained below with reference to a simplified example. Show this 1 . 2 a biaxial articulated robot 1 with two limbs 1.1 . 1.2 (see. 2 ), which are connected to each other via hinges whose joint angle q ₁ or q ₂ are adjustable via drives to move a load L whose inertia parameters, in particular mass m, center of gravity τ _{TCP, L} and inertial Torsor J _L respect of the Tool Center Points TCP.

In 1 the robot moves 1 the load L with folded arm 1.2 (q ₂ = 90 °) on a circular path with the Cartesian velocity v _{L, 1} counterclockwise upward (v <0), in 2 with folded arm 1.2 and higher angular velocity dq ₁ / dt down.

In order to protect operators and equipment as well as to avoid self-collisions, is in a world coordinate system, which is in the base of the robot 1 arranged and in the 1 . 2 dash-dotted lines, one in the figures cross-hatched indicated protection area A0, in which the robot 1 for example with its TCP or with none of its members 1.1 . 1.2 , allowed to penetrate. If, for example, the joint angle q = (q ₁ , q ₂ ) determines that the robot 1 Protected area A0 injured, it is braked and shut down safely.

To further penetration of the robot 1 to prevent the protection area A0 after it has violated its limit, this is increased on the basis of a maximum braking distance s _{max of} the robot. In 1 . 2 is exemplary of the circular segment-shaped protection area A0 at its two the permissible working space of the robot 1 facing radii (right in 1 . 2 ) symmetrically about a respective hatched circle segment A1 ( 1 ) or A2 ( 2 ) whose angular range is from the maximum braking distance s _{max, 1} or s _{max, 2} > s _{max, 1 of} the robot 1 depends. In another embodiment, not shown, the magnification takes place depending on the state of motion of the robot, for example, only on outer sides of the protection area whose outwardly directed normal with the speed of the TCP or a robot member includes an angle greater than 90 ° or which is a part of the Robot approaches, in 1 So just by adding the top, in 2 by adding the lower circle segment.

It can be seen from the simplified example that the maximum braking distance depends on the state of the robot: while its kinetic energy to be destroyed during deceleration 1 is lower. extends the braking distance in 2 due to the unfolded arm, the higher angular velocity dq ₁ / dt and the supporting gravitational effect.

For the braking distance s _max , therefore, during operation online depending on the state of the robot 1 according to a in 3 illustrated method estimated an upper bound.

To do this offline in advance for different types of robots, such as in 1 . 2 sketched biaxial robot with limbs 1.1 . 1.2 known length and mass distribution, each for a reference load L _Ref in a reference arrangement in a step S10 for different joint angle q _i and velocities dq _j / dt = (dq ₁ ^J / dt, dq ₂ ^J / dt, ...) based a dynamic model of the robot, preferably taking into account various Achsbremsmomentverläufe τ _l (q, dq / dt, t) or with a state of all variations, preferably minimal, fixed braking torque curve for each axis, respectively determined or estimated its maximum braking distance: s _{max, Ref, ij} = f (q _i , dq _j / dt, L _ref , τ _l ); i = 1, 2, ...; j = 1, 2, ...; l = 1, 2, ... (5)

Alternatively, these maximum braking distances can also be determined empirically.

These joint angular velocities dq _j / dt are also mapped to a scalar total velocity Q _j which is proportional to the sum of weighted powers of the individual joint angular velocities dq _k / dt (step S20): Q _j = a ₁ × (dq ₁ ^J / dt) ² + a ₂ × (dq ₂ ^J / dt) ² + ... = (dq ^j / dt) ^T × A × (dq ^j / dt) (6)

Then, in a step S30 of all simulated joint angular velocities dq / dt, which are mapped to the same overall velocity Q within a predetermined discretization δ, the respective maximum of the simulated braking distances s _{max, Ref, ij is} selected: F (x) = max {f (q _i , dq _j / dt, L _Ref ) | Q (dq _j / dt) = x ± δ} (7) where "max {Y (x) | Z (x) = X} "in a usual way the maximum of all values Y, whose argument x has the property Z (x) = X.

These maximum braking distances F related to the overall scalar velocity Q are now mapped to a polygon P in a step S40, which always has at least the maximum value of all maximum braking distances F in a discretization interval Q _k ≦ Q ≦ Q _{k + 1} . Q _k ≤ Q ≤ Q _{k + 1} ⇒ P (Q) ≥ F (Q). (8th)

This traverse thus provides an easily evaluable estimate for the maximum braking distance of a robot of the underlying type with a reference load L _Ref in a reference arrangement.

To estimate the maximum braking distance that occurs during operation, this value is still scaled from the reference values to the actual load and its arrangement. For this purpose, the value is multiplied on the one hand by a scaling factor fak _Last , which results from the quotient of the values according to (6) on the basis of actual load parameters divided by values according to (6) on the basis of the reference load parameters. In order to facilitate this evaluation, maximum values for the distances between the load center of gravity and the individual movement axes of the robot are selected, for example J _{y, L} + (1 _TCP + _1.1 + _1.2 ) ² × m _L with the moment of inertia J _{y, L} of Load around their center of gravity, the distance l _tcp from this to Flange of the robot, whose link lengths _1.1 , _1.2 and the mass m _{L of} the payload.

In addition, the value is multiplied by a scaling factor fak _TCP , which is the quotient of an auxiliary function for the geometric dimensions of the actual TCP divided by the same auxiliary function for the dimensions of a reference TCP, the auxiliary function itself being the one formed across all axes of the robot Sum of the quotients from the distance of the TCP to an axis, multiplied by its velocity square, divided by the angular acceleration of this axis.

Finally, in a step S50, simulations with different payloads and payload arrangements L _{k are} carried out offline and the maximum braking distances thus obtained are scaled back to the reference _load or arrangement by division with the above-described scaling factors fak _load , fak _TCP . If, for one or more overall speeds, this results in a higher value in comparison to the traverse determined on the basis of the reference load or arrangement, this is modified accordingly to P _mod , so that it at intervals in which the simulations with different loads from the reference configuration and payload arrangements give greater maximum braking distances, always having these higher values. Alternatively, the simulations with different payloads and payloads L _{k may} also be performed before, with or after steps S10, S20, and then lower further braking distances F (Q) related to the reference payload by the scaling back to those at step S40 in formation of the polygon P (Q).

In online operation, an upper limit for the maximum braking distance s _max is now estimated. For this purpose, in a cyclically repeated step S100, the current total speed Q for the current state is determined according to (5), with this from the modified polygonal path P _mod (Q) in step S110 determines a maximum reference braking and this in a step S120 with the scaling factors fak _Last , fak _TCP multiplied. On the basis of the thus estimated maximum braking distance s _max , the protected area A0 is widened in a step S130 in the manner explained in the introduction, wherein, for example, the angle of the widening circle segment A1 or A2 is selected to be proportional to the maximum braking distance s _max .

In the exemplary embodiment, due to the unfolded arm and the greater Cartesian velocity v _{L, 2} higher kinetic energy and the accelerating gravitational component of in 2 shown configuration to a larger estimated maximum braking distance s _max and thus to a larger magnification A2 of the protection area A0 than in the 1 illustrated configuration with folded arm and additional braking gravity component. This ensures that the robot comes to a standstill before the actual protection area A0 when it is in the state of 1 the upper limit of the range A1, in the state of 2 exceeds the lower limit of the range A2.

In this way, an upper limit for the maximum braking distance can be estimated in a computationally fast and simple manner and the protection range can be increased on the basis of this estimate.

LIST OF REFERENCE NUMBERS

1

robot

1.1, 1.2

robot members

A0, A1, A2

the scope

L

payload

s _max

maximum braking distance

q ₁ , q ₂

joint angle

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 102008013431 A1 [0004]

Claims (11)

Method for stopping a manipulator, in particular a robot ( 1 ), comprising the steps of: monitoring a range (A0; A1; A2); and braking the manipulator ( 1 ) in the case of a division violation; characterized in that the monitored area is variably set during operation on the basis of a braking distance (s _max ) of the manipulator (S130). Method for determining a braking distance, in particular for specifying the monitored area (A0, A1, A2) on the basis of the braking distance in a method according to claim 1, with the steps: pre-simulation of braking distances for different states (q, dq / dt, L) and / or braking force curves (τ) of the manipulator (S10); and estimating an upper limit as a braking distance (s _max ) based on the previously simulated braking distances (S120). A method according to claim 2, characterized in that braking paths for different states (q, dq / dt) and / or braking force curves (τ) of the manipulator in each case for a reference load and / or Referenztraglastanordnung (L _Ref ) are simulated in advance (S10). A method according to claim 3, characterized in that braking paths, which are simulated in advance for a reference load and / or Referenztraglastanordnung (L _Ref ), are scaled to an actual load and / or load assembly (L) (S120). Method according to one of the preceding claims 3 to 4, characterized in that braking paths for different states (q, dq / dt) and / or braking force curves (τ) of the manipulator in addition to the reference load and / or reference Traglastanordnung different payloads or load capacity arrangements (L _k ) are simulated in advance (S50). Method according to one of the preceding claims 2 to 5, characterized in that multidimensional states (dq / dt) of the manipulator are mapped to a lower-dimensional, in particular scalar quantity (Q), with respect to the braking distance is determined (S20, S100). Method according to one of the preceding claims 2 to 6, characterized in that the upper bound is determined on the basis of a polygon (P) (S40, S110). Method according to one of the preceding claims 2 to 7, characterized in that the braking distance is estimated on the basis of maximum values for load parameters. Device for stopping a manipulator, in particular a robot ( 1 ), with a computing device adapted to carry out a method according to any one of the preceding claims. Computer program that executes a method according to one of claims 1 to 8, when it runs in a computing device according to claim 9. A computer program product with program code stored on a machine readable medium and comprising a computer program according to claim 10.

Images